[NO PLAYERS] Seeding and ranking discussion

[GermanJoey]

Yeah, I'm definitely not feeling confident in the ratings. They were just the one way I could think of that we could both a.) incorporate a lot of people's feedback and b.) honestly make the ratings a reflection of everybody's thinking, and c.) make it simple enough so that everybody could understand it. But, nobody said anything about the formulas, and on top of that we just didn't get enough votes on anything. I mean, for the map, we have a sample size of 5? That's awful. For neighbor-strength, its only 4? Yeesh.

I personally feel you're over-estimating player skill vs map position (and besides, OH is really as much between a rock and a hard place as we could possibly put him, between gav and 2mn)... but, like, whether I agree or disagree shouldn't even matter as long as we have a way to make sure to include everybody's opinion into the result, e.g. everybody votes and then we tally up the average. "Lurkers try to balance players on a massively imbalanced, random map." We want everyone's opinion to matter somehow....

I could try to play with the formulas a bit... for example, instead of using player skill directly in the final balance formula, I could square each player's skill then divide by the average skill, then rescale so the result is still between 1 and 10. That would spread out the skill values some more. Or, if anyone else has a suggestion........................... suggestion... suggestion...

[GermanJoey]

I'm fine with GJ's last, but could someone through up a map with the layout as suggested by me above? Just for the sakes of considering an alternative, especially as 2.5 points seems like a lot to fix through combos alone.

I'll try out your setup too; I'll post about it later tonight. This one, right:

Regarding the new set-up, I would swap Molach and Greywolf, I think the large island position is stronger and more easily playable, even with Gavagai nearby -- as Gavagai will be very busy with OH. This applies even more after REM is brought into the centre, I think at this point Molach's start becomes strategically stellar.

Islands to the east of TW still need to be brought 2 tiles south, I think.


Frankly I don't think Gav-OH-2mn continent can be made to work, the disparity in geography is too great. Also, Gav and OH will probably expect to border each other, it's more fun to try and mess with them by having weaker, but agressive players like REM and 2mn next to them, plus weaker starts. But I'm not yet sure what the alternative suggestion would be (something like Gav in the centre of the big continent, REM to his south-east, Molach to his north-east, OH in Gav's current spot, 2mn in OH's current spot, ipecac in 2mn's current spot, Dantski in ipecac's current spot, Alhaz in Molach's current spot, Greywolf in REM's spot)

[AdrienIer]

I think removing the mountain S of player 2 is necessary so they can have early access to copper

[GermanJoey]

That's a good idea Adrien. IIRC, Commodore is going to balance strat resources and do a couple other small fixes (e.g. modifying player 7's start so they dont want to stab themselves in the gut on T1 when they find out they just orphaned three fish) once we finalize who is where on the map.

Speaking of, here's the numbers for Bacchus's scheme:



Unfortunately, I think the numbers are spread out too much here, so much that we'd need major map modifications to make this work. I think that it's too late for that, and that we should just go with Commodore's second suggestion.

Also, I played around a bit with the way neighbors are incorporated in the equations, and I had a good idea. What if we add a constant value (what I labeled the "per-neighbor-bias") to the final average neighbor rating? This is to correct for the fact that being in the middle with lots of neighbors, who each have a "danger rating" of 0.0, should be considered harder than having an isolated start. So, if a player had an average neighbor rating of -1 before, and the "per-neighbor-bias" was equal to 1, then now they are at 0; if they were at +1 before, now +2. Since we subtract out the average error at the end, this bias gets canceled out from players who have an average number of neighbors. However, it "punishes" players who have few-yet-skilled neighbors (e.g. OH/Gav) by saying their starts are easier since they have less players to deal with, but "rewards" players with many-but-unskilled neighbors (e.g. REM) by acknowledging their more complicated situation. Here's what we have now:



(compare to before)

This corrects a few positions that didn't feel right. For example, REM was at 2.5 before, saying his start was incredibly easy, but now he's just 0.5. OTOH, the TheWannabe's isolated start was adjusted by +1.5 towards easy. This feels much better, at least as far as neighbors go. The only start that really needs adjusting badly is Alhazard's, as you yourself said when you suggested this scheme.

As for OH... I dunno. I personally feel that skill-wise, OH being about 25% more skilled than Gavagai feels about right, I guess? I think I hold Gavagai in more esteem than you do... but, I also don't know what "25% better" actually means in specific game-terms. But, remember that these skill ratings of "10" and "8" don't really mean anything here beyond trying to match how good a player is compared to their position on the map. "10" and "8" aren't supposed to be reflections of people's standings in a site-wide tier list. In OH's case, his "10" just meant that he was assigned the spot on the map that the majority of lurkers rated the worst by a good margin:



And, because 10 is still higher than 8.3, he gets penalized further on top of that. He's also got the two "most dangerous" players by a good margin as close-by land neighbors.

In the end, he gets "average" combos according to this grid, which is I guess a boon for him, but then his close neighbor, 2metra+OT4E, get combos +2 tiers above average - specifically because of OH's large danger rating.

Also note that the map ratings have a huge amount of variance. Harry's start ranges from 7 to 9.5, while Gav's start ranges from 6 to 10! The map ratings do have a big effect on the outcome. For example, if the average for Harry's start were at 7 and Gav's at 10, then Gav would rate at -2 (too hard; would need land buffs) while Harry would be at +1.5 (too easy; gets combo picks 3 tiers below average). Likewise, if Harry's spot was rated at average 10 and Gav's 7, the situation would be reversed.

[Commodore]

Totally not disrespecting Gav, just amazingly respectful of Harry.

Yeah, part of OH's huge variance was probably due its extremely weird triple quarry-ness. I made it more conventional, here, and fixed the overland nature of The Wannabe's start.


This process won't get finished; we just need to pick when to shoot the engineers and let these guys play.

[Krill]

I almost regret that I haven't been wiped off the PB27 map yet, if only to be able to read about my mistakes in the other threads.

I'm sure someone is working on it.

[GermanJoey]

Totally not disrespecting Gav, just amazingly respectful of Harry.

Yeah, part of OH's huge variance was probably due its extremely weird triple quarry-ness. I made it more conventional, here, and fixed the overland nature of The Wannabe's start.


This process won't get finished; we just need to pick when to shoot the engineers and let these guys play.

I just looked at the changes, and I think they're pretty good. I liked the changes you made to Harry's start in particular; it's maybe about the same overall strength in isolation, but a lot less super-fast ancient-era which should make it a little tougher for him in ancient-era against his two strong neighbors. Also, the tradeoff of whether to take the clam vs the 2h plant (and permanently orphan the clam) should make him think quite a bit. The changes to player 8, 7, 2 and 6 all look good.

And I agree, lets just get this the hell done with. Maybe set a final deadline for map/combo discussion by tomorrow morning, unless anyone has any major objections?

[GermanJoey]

Ok, here's what the ratings said each player should get:


OH - mid tier combo
Gav - mid tier combo
2metra - combo from 2 tiers up
REM - combo from 1 tier down
Alhazard - combo from 2 tiers up
ipecac - combo from 1 tier down
Greywolf - combo from 1 tier up
The Wannabe - combo from 2 tiers down
Dantski - combo from 2 tiers up
Molach - combo from 2 tiers down

which translates into a tier list like this:

tier 1: 2metra, dantski, alhazard
tier 2: greywolf
tier 3: OH, Gav
tier 4: ipecac, REM
tier 5: molach, thewannabe

Here's the last posted combo list. The only one I have major disagreements with is Liz of Arabia, which I think is horrible, but the other two in that set are pretty good so whatever. Also, I just realized that Ramses of Khmer might be need to have an eye kept on it depending on what player gets it - they voted to ban War Elephants except for the Khmer UU.

(labels in white italics are mine)

Triplets of descending quality:

set 1 - (tier 1)

Victoria French
Cyrus Indian
Darius I Ottoman

set 2 - (tier 1)
Catherine Byzantine
Isabella Sumerian
Mansa Musa Babylonian

set 3 - (tier 1)
Julius Caesar Viking
Pericles Portuguese
Huayna Capac Russian

set 4 - (tier 2)
Elizabeth Arabian
Roosevelt Roman
Zara Yaqob Spanish

set 5 - (tier 3)
Louis XIV Dutch
Boudica Zulu
Kublai Khan Greek

set 6 - (tier 3)
Hatshepsut Incan
Justinian I Japanese
Frederick Egyptian

set 6 - (tier 4)
Ramesses II Khmer
Napoleon Native American
Gandhi Korean

set 7 - (tier 4)
Wang Kon German
Shaka Celtic
Qin Shi Huang Mali

set 8 - (tier 5)
Gilgamesh Aztec
Alexander Carthaginian
Stalin Mongolian

set 9 - (tier 5)
Hammurabi Persian
Charlemagne English
Lincoln Ethiopia

Several objectives; clearly these are ordered by quality but I'd like for there to also be some real choice in between the three options for each player.

So, we could pick any of the T1 civs to give any of the T1 sets to. I don't have time right now to check how well any of the sets matches any of the combos. (halloween party) I'll take a look, as well as do a check for strat resource coverage, assuming I'm still conscious, when I get home tonight.

[Commodore]

If you want to do it thus, I'd use experience as the "rank in-tier":

tier 1: 2metra, dantski, alhazard
tier 2: greywolf
tier 3: OH, Gav
tier 4: ipecac, REM
tier 5: molach, thewannabe

So....

Triplets of descending quality:
Dantski
set 1 - (tier 1)

Victoria French
Cyrus Indian
Darius I Ottoman

Alhazard
set 2 - (tier 1)
Catherine Byzantine
Isabella Sumerian
Mansa Musa Babylonian

2metraninja
set 3 - (tier 1)
Julius Caesar Viking
Pericles Portuguese
Huayna Capac Russian

Greywolf
set 4 - (tier 2)
Elizabeth Arabian Ragnar English
Roosevelt Roman
Zara Yaqob Spanish

Gavagai
set 5 - (tier 3)
Louis XIV Dutch
Boudica Zulu
Kublai Khan Greek

Old Harry
set 6 - (tier 3)
Hatshepsut Incan
Justinian I Japanese
Frederick Egyptian

Ipecac
set 6 - (tier 4)
Ramesses II Khmer
Napoleon Native American
Gandhi Korean

R.E.M.
set 7 - (tier 4)
Wang Kon German
Shaka Celtic
Qin Shi Huang Mali

Molach
set 8 - (tier 5)
Gilgamesh Aztec
Alexander Carthaginian
Stalin Mongolian

TheWannabe
set 9 - (tier 5)
Hammurabi Persian
Charlemagne English Elizabeth Arabian
Lincoln Ethiopia

Only person I'd feel really bad for in this would be poor, poor Wannabe. Maybe give him Elizabeth of Arabia if you rate it that low? Really want to give the dude Financial option.

[GermanJoey]

Sure. Pick out whatever one you think is weakest from TheWannabe (charley?) and then give Greywolf someone similarish to Liz of Bad Techs. Uh, I dunno, Ragnar of someone? Maybe Freddy? (already the choice of someone else) of England? Or, is there any other civ that isn't in the list?